Accurate extended time horizon drought prediction via wavelet-season-fuzzy models

Abstract

Improved accurate prediction of the Palmer drought severity index (PDSI) is crucial for sustainable water supply demand, flood mitigation, management of hydraulic structures, sustainability of the ecosystem and, significant economic and social benefits. In this study, a novel predictive ASA-fuzzy model based on additive season algorithm (ASA) and fuzzy logic is proposed to enhance prediction accuracy with extended future lead times. For the first time, additive season algorithm (ASA) is introduced as an alternative data preprocessing algorithm in prediction of PDSI data obtained from measurement stations that exhibit distinct meteorological characteristics. The results showed that the newly proposed hybrid ASA-fuzzy approach can satisfactorily be utilized to predict monthly PDSI data up to 24-month time horizon. In comparison, for all stations’ data, the newly proposed ASA-fuzzy is found to be superior in accuracy to the stand-alone fuzzy and widely used W-fuzzy models for all lead time predictions based on the quantitative diagnostic measures, root mean squared error (RMSE) and the Nash–Sutcliffe coefficient of efficiency (CE). The remarkable performance of the introduced ASA-fuzzy model in this study clearly shows that the season algorithm can decompose the original data into trend cycle, seasonality and error components more effectively than wavelet technique. Therefore ASA-Fuzzy model has been advocated as a new prediction tool in predicting monthly PDSI data with superior accuracy when compared to conventional methods. In addition, when hybridized with a proper preprocessing algorithm, power of the fuzzy modeling approach in time-series forecasting is manifested.